Discussion Speculation: Zen 4 (EPYC 4 "Genoa", Ryzen 7000, etc.)

[Vattila]

Except for the details about the improvements in the microarchitecture, we now know pretty well what to expect with Zen 3.

The leaked presentation by AMD Senior Manager Martin Hilgeman shows that EPYC 3 "Milan" will, as promised and expected, reuse the current platform (SP3), and the system architecture and packaging looks to be the same, with the same 9-die chiplet design and the same maximum core and thread-count (no SMT-4, contrary to rumour). The biggest change revealed so far is the enlargement of the compute complex from 4 cores to 8 cores, all sharing a larger L3 cache ("32+ MB", likely to double to 64 MB, I think).

Hilgeman's slides did also show that EPYC 4 "Genoa" is in the definition phase (or was at the time of the presentation in September, at least), and will come with a new platform (SP5), with new memory support (likely DDR5).

What else do you think we will see with Zen 4? PCI-Express 5 support? Increased core-count? 4-way SMT? New packaging (interposer, 2.5D, 3D)? Integrated memory on package (HBM)?

Vote in the poll and share your thoughts!

 

[leoneazzurro]

It has a yellow color tone to it. Perhaps Titanium Dioxide layer behind the super thin Silver layer?

Silicon Oxide too has a yellowish color in certain finishes.

 

[Timorous]

That there is quite probably headroom to increase clock even more (at the expense of power budget, of course). That is, he's hinting at easy overclock.

Don't know about easy overclock. I assume they are stull tuning the boost algorithms so the prototype was running conservatively and still hitting 5.5Ghz. I expect with more tweaking on the boost algs and with more info on how parts bin they will get that dialed in so there won't be much left to overclock.

 

[gdansk]

If Robert says 15% ST we can easily expect 14.2%
(I'm half joking, remember the Zen 2 boost debacle)

 

[Hitman928]

That there is quite probably headroom to increase clock even more (at the expense of power budget, of course). That is, he's hinting at easy overclock.

I know this is what he was implying, I just don't trust marketing folk, especially with vague, unqualified statements like this. I highly doubt the engineers thought hitting 5.5 GHz was easy. Anyway, we'll see how much room is left above that. There's obviously some room for the best cores as they hit 5.5 GHz while playing a game that uses at least 4 cores effectively.

 

[uzzi38]

If Robert says 15% ST we can easily expect 14.2%
(I'm half joking, remember the Zen 2 boost debacle)

Trust me, you won't have to worry about frequencies this generation lmfao

 

[nicalandia]

Silicon Oxide too has a yellowish color in certain finishes.

It's the Nickel Vanadium(NiV) layer that sits below the Silver Layer.

Both Titanium and Silicon Oxide are a poor Thermal/Electrical conductors. It will be counterintuitive to what backside Metallization is trying to achieve.

 

[tomatosummit]

My favourite from that is
""With 28 Gen 5 lanes from the CPU, 16 PEG lanes, and 4 chipset lanes, 8 lanes are left. Is it possible for motherboards to have two M.2 slots wired to the processor?
Yes, that is a possibility. ""
I'd hope that's the default for b650 boards, having high speed storage sharing the pch uplink with everything else has always rubbed me the wrong way.

Are those extra four lanes for the second chipset in 670boards? he doesn't imply that and the leaked board looked like it had a huge switch in the middle.

 

[Saylick]

Trust me, you won't have to worry about frequencies this generation lmfao

I get the impression we might see 6 GHz single core with manual OC, power consumption be damned!

 

[inf64]

I get the impression we might see 6 GHz single core with manual OC, power consumption be damned!

Based on that TPU interview, it looks like 6Ghz manual OC is pretty much a given (if they get 5.5Ghz on all cores in games). I think that with PBO and proper cooling, 5.5Ghz all core boost will be doable and easy to achieve. Couple that with at least 10% IPC (I hope it will be more than that) and high DDR5 speeds, this thing will be a dominant high end desktop chip.

 

[leoneazzurro]

It's the Nickel Vanadium(NiV) layer that sits below the Silver Layer.

Both Titanium and Silicon Oxide are a poor Thermal/Electrical conductors. It will be counterintuitive to what backside Metallization is trying to achieve.

But I don't see a silver layer there, at least at what I am referring to, that is the top of the CCDs Lisa had in hand.

 

[Karnak]

I get the impression we might see 6 GHz single core with manual OC, power consumption be damned!

Well if it's only single core then power consumption doesn't matter at all.

 

[coercitiv]

I get the impression we might see 6 GHz single core with manual OC, power consumption be damned!

Oh don't you worry, with 230W PPT AMD will overclock it for you! You just sit back, relax, get those fans in push-pull configuration!

 

[Anhiel]

Yeah, if the engineering sample was indeed a 125W TDP/170PPT then even 6.6-6.8GHz might be possible for OC.

Anyhow, talking about efficiency so I pulled up AMD's official propaganda on their 25x20, 30x25 projects and their slide on Zen-Zen3 efficiency. So based on their data Zen4 needs to hit 4.6x compared to Zen to meet the estimated mean curve. (I censored anything not official)

 

[Saylick]

Based on that TPU interview, it looks like 6Ghz manual OC is pretty much a given (if they get 5.5Ghz on all cores in games). I think that with PBO and proper cooling, 5.5Ghz all core boost will be doable and easy to achieve. Couple that with at least 10% IPC (I hope it will be more than that) and high DDR5 speeds, this thing will be a dominant high end desktop chip.

Yeah, my takeaway from that interview was that we probably can see 5.6, maybe 5.7 GHz single core turbo out of the box. It hopefully wouldn't be too difficult to squeeze out another few hundred MHz with a manual OC on top of that. I'm not convinced that all of the cores can hit 5.5 GHz for most workloads simply because that game isn't heavily multi-threaded apparently, but at least it showed that at least the cores themselves can hit high clocks if power/thermals aren't an issue.

 

[Exist50]

I'm thinking at least another couple hundred MHz should be possible, and 6GHz overclocks a distinct possibility. Probably not much beyond that, but N5P (and related) should give a really solid frequency boost. Hah, at this rate, maybe we'll see 7GHz by N2.

 

[Saylick]

Well if it's only single core then power consumption doesn't matter at all.

Netburst says hi.

 

[Thunder 57]

I see the hype train is running full speed again. Choo Choo! I for one do not expect 6GHz overclocks, maybe 5.7-5.8. But in the words of principal Skinner, "Prove me wrong kids, prove me wrong!". For clarification that was a response to "Come on Edna, we both know these kids have no future", for those that would otherwise not get the reference.

Edited for clarity.

 

[nicalandia]

But I don't see a silver layer there, at least at what I am referring to, that is the top of the CCDs Lisa had in hand.

It would appear that very thin layer of silver will look yellow due to a phenomenon known as thin-film interference

 

[jamescox]

Well, we don't know if it was using 5 W less or 50 W less so it's a reply that doesn't actually clarify much other than it was almost assuredly above 170 W. Either way, I don't expect performance to increase much even though the CPU was using less than it's full power target as you are so far into the inefficient part of the curve that small increases in performance take overly large amounts of power increases. That is, unless AMD marketing is feigning incompetence to hardcore sandbag everyone but I really doubt this is the case. Don't get me wrong, Zen 4 will still be a very strong CPU and most likely the best money can buy in almost every way upon release and potentially even after RPL drops, but I would rather see them use their node advantage to really work on IPC and keeping power efficiency as the dominant feature over their competition, rather than follow Intel with ever faster, more power hungry cores.

I doubt that these are being pushed that high up the frequency / power curve, at least compared to what Intel has been doing. This may not even be for any Zen 4 processor. It might be future expandability for Zen 5. You obviously get less efficient with higher clocks, but if it delivers on performance without going to power extremes, then that is fine. If you don’t like it, then lock it at the base clock or something. If they can run over 5 GHz all core, I think a lot of users of multi-threaded applications will be okay with some extra power usage. People would also be fine with a high end version that requires extreme cooling anyway.

I am still wondering about the thick lid. Such a thick piece of metal generally isn’t good for thermal conductivity. The thermal conductivity of heat pipes can be orders of magnitude higher than any solid material (it is essentially passive phase change), so it is best to get the heat pipes as close to the die as possible.

 

[moinmoin]

Are those extra four lanes for the second chipset in 670boards?

No, the second chipset is daisy-chained to the first chipset.

Yeah, if the engineering sample was indeed a 125W TDP/170PPT then even 6.6-6.8GHz might be possible for OC.

Well, that escalated quickly.

 

[Saylick]

I am still wondering about the thick lid. Such a thick piece of metal generally isn’t good for thermal conductivity. The thermal conductivity of heat pipes can be orders of magnitude higher than any solid material (it is essentially passive phase change), so it is best to get the heat pipes as close to the die as possible.

If you thought the heat spreader for AM4 was thick, then I guess you can call this one thick too, but the z-height of the IHS is the same as for AM4, so if the z-height of the dies are the same, chances are the lid is the same thickness as before.

Why the new heatspreader design? Why the holes on the sides?
It's actually how we achieve cooler compatibility. If you flip over one of the AM4 processors, you'll find a blank spot in the middle without pins, which has space for capacitors. That blank space is not available on Socket AM5, it has LGA pads across the entire bottom surface of the chip. We had to move those capacitors somewhere else. They don't go under the heatspreader due to thermal challenges, so we had to put them on top of the package, which required us to make cutouts on the IHS to make room. Because of those changes we're able to keep the same package size, length and width, same z-height, same socket keep-out pattern, and that's what enables cooler compatibility with AM4.

 

[Schmide]

I am still wondering about the thick lid. Such a thick piece of metal generally isn’t good for thermal conductivity. The thermal conductivity of heat pipes can be orders of magnitude higher than any solid material (it is essentially passive phase change), so it is best to get the heat pipes as close to the die as possible.

I think in these days with such extreme power management, the larger thermal capacity and improved integrated thermal interface material far outclass the ability of the average user to improve upon the thermal resistance (c/w) of the system. The days of deliding have really fell on diminished returns.

 

[Markfw]

If you thought the heat spreader for AM4 was thick, then I guess you can call this one thick too, but the z-height of the IHS is the same as for AM4, so if the z-height of the dies are the same, chances are the lid is the same thickness as before.

The socket for AM4 with the pins is pretty high compared to a LGA socket, so the extra height I think is as said to make the same z-height. Also the way its made, it will be more "bend-resistant" being thick, since the sides are not all there. I think its pretty ingenious.

 

[leoneazzurro]

It would appear that very thin layer of silver will look yellow due to a phenomenon known as thin-film interference

Well I see silver layers as thin as 50nm everyday at work on top of Nickel or NiV layers and they still appear grayish to me. But OK, in the interview they say explicitely it's their backside metallization.

 

[Hitman928]

I doubt that these are being pushed that high up the frequency / power curve, at least compared to what Intel has been doing. This may not even be for any Zen 4 processor. It might be future expandability for Zen 5. You obviously get less efficient with higher clocks, but if it delivers on performance without going to power extremes, then that is fine. If you don’t like it, then lock it at the base clock or something. If they can run over 5 GHz all core, I think a lot of users of multi-threaded applications will be okay with some extra power usage. People would also be fine with a high end version that requires extreme cooling anyway.

I am still wondering about the thick lid. Such a thick piece of metal generally isn’t good for thermal conductivity. The thermal conductivity of heat pipes can be orders of magnitude higher than any solid material (it is essentially passive phase change), so it is best to get the heat pipes as close to the die as possible.

Robert Hallock confirmed both that the Ryzen 7000 series would be using up to 230 W PPT and that the CPU they demoed was in the 170 W TDP /230 W PPT class.

AMD is free to do and design products that they think are best and I hope they continue to be very successful, I just wished there was a bigger push for efficiency is all. I probably won't upgrade to Zen 4 anyway as I'm pretty happy with Zen 3 so we'll see what Zen 5 / whatever Intel has for desktops at that point brings.